Go to ICLR 2025 Conference homepageProcedural Fairness Through Addressing Social Determinants of Opportunity
Social determinants of opportunity are variables that, while not directly pertaining to any specific individual, capture key aspects of contexts and environments that have direct causal influences on certain attributes of an individual, e.g., environmental pollution in an area affects individual's health condition, and educational resources in an neighborhood influence individual's academic preparedness. Previous algorithmic fairness literature often overlooks social determinants of opportunity, leading to implications for procedural fairness and structural justice that are incomplete and potentially even inaccurate. We propose a modeling framework that explicitly incorporates social determinants of opportunity and their causal influences on individual-level attributes of interest. To demonstrate theoretical perspectives and practical applicability of our framework, we consider college admissions as a running example. Specifically, for three mainstream admission procedures that have historically been implemented or are still in use today, we distinguish and draw connections between the outcome of admission decision-making and the underlying distribution of academic preparedness in the applicant population. Our findings suggest that mitigation strategies centering solely around protected features may introduce new procedural unfairness when addressing existing discrimination. Considering both individual-level attributes and social determinants of opportunity facilitates a more comprehensive explication of benefits and burdens experienced by individuals from diverse demographic backgrounds as well as contextual environments, which is essential for understanding and achieving procedural fairness effectively and transparently.